1 // SPDX-License-Identifier: GPL-2.0-or-later <<
2 /* SCTP kernel implementation 1 /* SCTP kernel implementation
3 * (C) Copyright IBM Corp. 2003, 2004 2 * (C) Copyright IBM Corp. 2003, 2004
4 * 3 *
5 * This file is part of the SCTP kernel implem 4 * This file is part of the SCTP kernel implementation
6 * 5 *
7 * This file contains the code relating the ch 6 * This file contains the code relating the chunk abstraction.
8 * 7 *
>> 8 * This SCTP implementation is free software;
>> 9 * you can redistribute it and/or modify it under the terms of
>> 10 * the GNU General Public License as published by
>> 11 * the Free Software Foundation; either version 2, or (at your option)
>> 12 * any later version.
>> 13 *
>> 14 * This SCTP implementation is distributed in the hope that it
>> 15 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
>> 16 * ************************
>> 17 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
>> 18 * See the GNU General Public License for more details.
>> 19 *
>> 20 * You should have received a copy of the GNU General Public License
>> 21 * along with GNU CC; see the file COPYING. If not, see
>> 22 * <http://www.gnu.org/licenses/>.
>> 23 *
9 * Please send any bug reports or fixes you ma 24 * Please send any bug reports or fixes you make to the
10 * email address(es): 25 * email address(es):
11 * lksctp developers <linux-sctp@vger.kerne 26 * lksctp developers <linux-sctp@vger.kernel.org>
12 * 27 *
13 * Written or modified by: 28 * Written or modified by:
14 * Jon Grimm <jgrimm@us.ibm.com 29 * Jon Grimm <jgrimm@us.ibm.com>
15 * Sridhar Samudrala <sri@us.ibm.com> 30 * Sridhar Samudrala <sri@us.ibm.com>
16 */ 31 */
17 32
18 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 33 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
19 34
20 #include <linux/types.h> 35 #include <linux/types.h>
21 #include <linux/kernel.h> 36 #include <linux/kernel.h>
22 #include <linux/net.h> 37 #include <linux/net.h>
23 #include <linux/inet.h> 38 #include <linux/inet.h>
24 #include <linux/skbuff.h> 39 #include <linux/skbuff.h>
25 #include <linux/slab.h> 40 #include <linux/slab.h>
26 #include <net/sock.h> 41 #include <net/sock.h>
27 #include <net/sctp/sctp.h> 42 #include <net/sctp/sctp.h>
28 #include <net/sctp/sm.h> 43 #include <net/sctp/sm.h>
29 44
30 /* This file is mostly in anticipation of futu 45 /* This file is mostly in anticipation of future work, but initially
31 * populate with fragment tracking for an outb 46 * populate with fragment tracking for an outbound message.
32 */ 47 */
33 48
34 /* Initialize datamsg from memory. */ 49 /* Initialize datamsg from memory. */
35 static void sctp_datamsg_init(struct sctp_data 50 static void sctp_datamsg_init(struct sctp_datamsg *msg)
36 { 51 {
37 refcount_set(&msg->refcnt, 1); !! 52 atomic_set(&msg->refcnt, 1);
38 msg->send_failed = 0; 53 msg->send_failed = 0;
39 msg->send_error = 0; 54 msg->send_error = 0;
40 msg->can_delay = 1; 55 msg->can_delay = 1;
41 msg->abandoned = 0; <<
42 msg->expires_at = 0; 56 msg->expires_at = 0;
43 INIT_LIST_HEAD(&msg->chunks); 57 INIT_LIST_HEAD(&msg->chunks);
44 } 58 }
45 59
46 /* Allocate and initialize datamsg. */ 60 /* Allocate and initialize datamsg. */
47 static struct sctp_datamsg *sctp_datamsg_new(g 61 static struct sctp_datamsg *sctp_datamsg_new(gfp_t gfp)
48 { 62 {
49 struct sctp_datamsg *msg; 63 struct sctp_datamsg *msg;
50 msg = kmalloc(sizeof(struct sctp_datam 64 msg = kmalloc(sizeof(struct sctp_datamsg), gfp);
51 if (msg) { 65 if (msg) {
52 sctp_datamsg_init(msg); 66 sctp_datamsg_init(msg);
53 SCTP_DBG_OBJCNT_INC(datamsg); 67 SCTP_DBG_OBJCNT_INC(datamsg);
54 } 68 }
55 return msg; 69 return msg;
56 } 70 }
57 71
58 void sctp_datamsg_free(struct sctp_datamsg *ms 72 void sctp_datamsg_free(struct sctp_datamsg *msg)
59 { 73 {
60 struct sctp_chunk *chunk; 74 struct sctp_chunk *chunk;
61 75
62 /* This doesn't have to be a _safe vai 76 /* This doesn't have to be a _safe vairant because
63 * sctp_chunk_free() only drops the re 77 * sctp_chunk_free() only drops the refs.
64 */ 78 */
65 list_for_each_entry(chunk, &msg->chunk 79 list_for_each_entry(chunk, &msg->chunks, frag_list)
66 sctp_chunk_free(chunk); 80 sctp_chunk_free(chunk);
67 81
68 sctp_datamsg_put(msg); 82 sctp_datamsg_put(msg);
69 } 83 }
70 84
71 /* Final destructruction of datamsg memory. */ 85 /* Final destructruction of datamsg memory. */
72 static void sctp_datamsg_destroy(struct sctp_d 86 static void sctp_datamsg_destroy(struct sctp_datamsg *msg)
73 { 87 {
74 struct sctp_association *asoc = NULL; <<
75 struct list_head *pos, *temp; 88 struct list_head *pos, *temp;
76 struct sctp_chunk *chunk; 89 struct sctp_chunk *chunk;
>> 90 struct sctp_sock *sp;
77 struct sctp_ulpevent *ev; 91 struct sctp_ulpevent *ev;
78 int error, sent; !! 92 struct sctp_association *asoc = NULL;
>> 93 int error = 0, notify;
>> 94
>> 95 /* If we failed, we may need to notify. */
>> 96 notify = msg->send_failed ? -1 : 0;
79 97
80 /* Release all references. */ 98 /* Release all references. */
81 list_for_each_safe(pos, temp, &msg->ch 99 list_for_each_safe(pos, temp, &msg->chunks) {
82 list_del_init(pos); 100 list_del_init(pos);
83 chunk = list_entry(pos, struct 101 chunk = list_entry(pos, struct sctp_chunk, frag_list);
84 !! 102 /* Check whether we _really_ need to notify. */
85 if (!msg->send_failed) { !! 103 if (notify < 0) {
86 sctp_chunk_put(chunk); !! 104 asoc = chunk->asoc;
87 continue; !! 105 if (msg->send_error)
>> 106 error = msg->send_error;
>> 107 else
>> 108 error = asoc->outqueue.error;
>> 109
>> 110 sp = sctp_sk(asoc->base.sk);
>> 111 notify = sctp_ulpevent_type_enabled(SCTP_SEND_FAILED,
>> 112 &sp->subscribe);
88 } 113 }
89 114
90 asoc = chunk->asoc; !! 115 /* Generate a SEND FAILED event only if enabled. */
91 error = msg->send_error ?: aso !! 116 if (notify > 0) {
92 sent = chunk->has_tsn ? SCTP_D !! 117 int sent;
>> 118 if (chunk->has_tsn)
>> 119 sent = SCTP_DATA_SENT;
>> 120 else
>> 121 sent = SCTP_DATA_UNSENT;
93 122
94 if (sctp_ulpevent_type_enabled <<
95 <<
96 ev = sctp_ulpevent_mak 123 ev = sctp_ulpevent_make_send_failed(asoc, chunk, sent,
97 124 error, GFP_ATOMIC);
98 if (ev) 125 if (ev)
99 asoc->stream.s !! 126 sctp_ulpq_tail_event(&asoc->ulpq, ev);
100 } <<
101 <<
102 if (sctp_ulpevent_type_enabled <<
103 <<
104 ev = sctp_ulpevent_mak <<
105 <<
106 <<
107 if (ev) <<
108 asoc->stream.s <<
109 } 127 }
110 128
111 sctp_chunk_put(chunk); 129 sctp_chunk_put(chunk);
112 } 130 }
113 131
114 SCTP_DBG_OBJCNT_DEC(datamsg); 132 SCTP_DBG_OBJCNT_DEC(datamsg);
115 kfree(msg); 133 kfree(msg);
116 } 134 }
117 135
118 /* Hold a reference. */ 136 /* Hold a reference. */
119 static void sctp_datamsg_hold(struct sctp_data 137 static void sctp_datamsg_hold(struct sctp_datamsg *msg)
120 { 138 {
121 refcount_inc(&msg->refcnt); !! 139 atomic_inc(&msg->refcnt);
122 } 140 }
123 141
124 /* Release a reference. */ 142 /* Release a reference. */
125 void sctp_datamsg_put(struct sctp_datamsg *msg 143 void sctp_datamsg_put(struct sctp_datamsg *msg)
126 { 144 {
127 if (refcount_dec_and_test(&msg->refcnt !! 145 if (atomic_dec_and_test(&msg->refcnt))
128 sctp_datamsg_destroy(msg); 146 sctp_datamsg_destroy(msg);
129 } 147 }
130 148
131 /* Assign a chunk to this datamsg. */ 149 /* Assign a chunk to this datamsg. */
132 static void sctp_datamsg_assign(struct sctp_da 150 static void sctp_datamsg_assign(struct sctp_datamsg *msg, struct sctp_chunk *chunk)
133 { 151 {
134 sctp_datamsg_hold(msg); 152 sctp_datamsg_hold(msg);
135 chunk->msg = msg; 153 chunk->msg = msg;
136 } 154 }
137 155
138 156
139 /* A data chunk can have a maximum payload of 157 /* A data chunk can have a maximum payload of (2^16 - 20). Break
140 * down any such message into smaller chunks. 158 * down any such message into smaller chunks. Opportunistically, fragment
141 * the chunks down to the current MTU constrai 159 * the chunks down to the current MTU constraints. We may get refragmented
142 * later if the PMTU changes, but it is _much 160 * later if the PMTU changes, but it is _much better_ to fragment immediately
143 * with a reasonable guess than always doing o 161 * with a reasonable guess than always doing our fragmentation on the
144 * soft-interrupt. 162 * soft-interrupt.
145 */ 163 */
146 struct sctp_datamsg *sctp_datamsg_from_user(st 164 struct sctp_datamsg *sctp_datamsg_from_user(struct sctp_association *asoc,
147 st 165 struct sctp_sndrcvinfo *sinfo,
148 st 166 struct iov_iter *from)
149 { 167 {
150 size_t len, first_len, max_data, remai 168 size_t len, first_len, max_data, remaining;
151 size_t msg_len = iov_iter_count(from); 169 size_t msg_len = iov_iter_count(from);
152 struct sctp_shared_key *shkey = NULL; <<
153 struct list_head *pos, *temp; 170 struct list_head *pos, *temp;
154 struct sctp_chunk *chunk; 171 struct sctp_chunk *chunk;
155 struct sctp_datamsg *msg; 172 struct sctp_datamsg *msg;
156 int err; 173 int err;
157 174
158 msg = sctp_datamsg_new(GFP_KERNEL); 175 msg = sctp_datamsg_new(GFP_KERNEL);
159 if (!msg) 176 if (!msg)
160 return ERR_PTR(-ENOMEM); 177 return ERR_PTR(-ENOMEM);
161 178
162 /* Note: Calculate this outside of the 179 /* Note: Calculate this outside of the loop, so that all fragments
163 * have the same expiration. 180 * have the same expiration.
164 */ 181 */
165 if (asoc->peer.prsctp_capable && sinfo 182 if (asoc->peer.prsctp_capable && sinfo->sinfo_timetolive &&
166 (SCTP_PR_TTL_ENABLED(sinfo->sinfo_ 183 (SCTP_PR_TTL_ENABLED(sinfo->sinfo_flags) ||
167 !SCTP_PR_POLICY(sinfo->sinfo_flag 184 !SCTP_PR_POLICY(sinfo->sinfo_flags)))
168 msg->expires_at = jiffies + 185 msg->expires_at = jiffies +
169 msecs_to_jif 186 msecs_to_jiffies(sinfo->sinfo_timetolive);
170 187
171 /* This is the biggest possible DATA c 188 /* This is the biggest possible DATA chunk that can fit into
172 * the packet 189 * the packet
173 */ 190 */
174 max_data = asoc->frag_point; !! 191 max_data = asoc->pathmtu -
175 if (unlikely(!max_data)) { !! 192 sctp_sk(asoc->base.sk)->pf->af->net_header_len -
176 max_data = sctp_min_frag_point !! 193 sizeof(struct sctphdr) - sizeof(struct sctp_data_chunk);
177 !! 194 max_data = SCTP_TRUNC4(max_data);
178 pr_warn_ratelimited("%s: asoc: <<
179 __func__, <<
180 } <<
181 195
182 /* If the peer requested that we authe !! 196 /* If the the peer requested that we authenticate DATA chunks
183 * we need to account for bundling of 197 * we need to account for bundling of the AUTH chunks along with
184 * DATA. 198 * DATA.
185 */ 199 */
186 if (sctp_auth_send_cid(SCTP_CID_DATA, 200 if (sctp_auth_send_cid(SCTP_CID_DATA, asoc)) {
187 struct sctp_hmac *hmac_desc = 201 struct sctp_hmac *hmac_desc = sctp_auth_asoc_get_hmac(asoc);
188 202
189 if (hmac_desc) 203 if (hmac_desc)
190 max_data -= SCTP_PAD4( !! 204 max_data -= SCTP_PAD4(sizeof(sctp_auth_chunk_t) +
191 205 hmac_desc->hmac_len);
192 <<
193 if (sinfo->sinfo_tsn && <<
194 sinfo->sinfo_ssn != asoc-> <<
195 shkey = sctp_auth_get_ <<
196 if (!shkey) { <<
197 err = -EINVAL; <<
198 goto errout; <<
199 } <<
200 } else { <<
201 shkey = asoc->shkey; <<
202 } <<
203 } 206 }
204 207
>> 208 /* Check what's our max considering the above */
>> 209 max_data = min_t(size_t, max_data, asoc->frag_point);
>> 210
205 /* Set first_len and then account for 211 /* Set first_len and then account for possible bundles on first frag */
206 first_len = max_data; 212 first_len = max_data;
207 213
208 /* Check to see if we have a pending S 214 /* Check to see if we have a pending SACK and try to let it be bundled
209 * with this message. Do this if we d 215 * with this message. Do this if we don't have any data queued already.
210 * To check that, look at out_qlen and 216 * To check that, look at out_qlen and retransmit list.
211 * NOTE: we will not reduce to account 217 * NOTE: we will not reduce to account for SACK, if the message would
212 * not have been fragmented. 218 * not have been fragmented.
213 */ 219 */
214 if (timer_pending(&asoc->timers[SCTP_E 220 if (timer_pending(&asoc->timers[SCTP_EVENT_TIMEOUT_SACK]) &&
215 asoc->outqueue.out_qlen == 0 && 221 asoc->outqueue.out_qlen == 0 &&
216 list_empty(&asoc->outqueue.retrans 222 list_empty(&asoc->outqueue.retransmit) &&
217 msg_len > max_data) 223 msg_len > max_data)
218 first_len -= SCTP_PAD4(sizeof( !! 224 first_len -= SCTP_PAD4(sizeof(sctp_sack_chunk_t));
219 225
220 /* Encourage Cookie-ECHO bundling. */ 226 /* Encourage Cookie-ECHO bundling. */
221 if (asoc->state < SCTP_STATE_COOKIE_EC 227 if (asoc->state < SCTP_STATE_COOKIE_ECHOED)
222 first_len -= SCTP_ARBITRARY_CO 228 first_len -= SCTP_ARBITRARY_COOKIE_ECHO_LEN;
223 229
224 /* Account for a different sized first 230 /* Account for a different sized first fragment */
225 if (msg_len >= first_len) { 231 if (msg_len >= first_len) {
226 msg->can_delay = 0; 232 msg->can_delay = 0;
227 if (msg_len > first_len) !! 233 SCTP_INC_STATS(sock_net(asoc->base.sk), SCTP_MIB_FRAGUSRMSGS);
228 SCTP_INC_STATS(asoc->b <<
229 SCTP_MI <<
230 } else { 234 } else {
231 /* Which may be the only one.. 235 /* Which may be the only one... */
232 first_len = msg_len; 236 first_len = msg_len;
233 } 237 }
234 238
235 /* Create chunks for all DATA chunks. 239 /* Create chunks for all DATA chunks. */
236 for (remaining = msg_len; remaining; r 240 for (remaining = msg_len; remaining; remaining -= len) {
237 u8 frag = SCTP_DATA_MIDDLE_FRA 241 u8 frag = SCTP_DATA_MIDDLE_FRAG;
238 242
239 if (remaining == msg_len) { 243 if (remaining == msg_len) {
240 /* First frag, which m 244 /* First frag, which may also be the last */
241 frag |= SCTP_DATA_FIRS 245 frag |= SCTP_DATA_FIRST_FRAG;
242 len = first_len; 246 len = first_len;
243 } else { 247 } else {
244 /* Middle frags */ 248 /* Middle frags */
245 len = max_data; 249 len = max_data;
246 } 250 }
247 251
248 if (len >= remaining) { 252 if (len >= remaining) {
249 /* Last frag, which ma 253 /* Last frag, which may also be the first */
250 len = remaining; 254 len = remaining;
251 frag |= SCTP_DATA_LAST 255 frag |= SCTP_DATA_LAST_FRAG;
252 256
253 /* The application req 257 /* The application requests to set the I-bit of the
254 * last DATA chunk of 258 * last DATA chunk of a user message when providing
255 * the user message to 259 * the user message to the SCTP implementation.
256 */ 260 */
257 if ((sinfo->sinfo_flag 261 if ((sinfo->sinfo_flags & SCTP_EOF) ||
258 (sinfo->sinfo_flag 262 (sinfo->sinfo_flags & SCTP_SACK_IMMEDIATELY))
259 frag |= SCTP_D 263 frag |= SCTP_DATA_SACK_IMM;
260 } 264 }
261 265
262 chunk = asoc->stream.si->make_ !! 266 chunk = sctp_make_datafrag_empty(asoc, sinfo, len, frag,
263 !! 267 0, GFP_KERNEL);
264 if (!chunk) { 268 if (!chunk) {
265 err = -ENOMEM; 269 err = -ENOMEM;
266 goto errout; 270 goto errout;
267 } 271 }
268 272
269 err = sctp_user_addto_chunk(ch 273 err = sctp_user_addto_chunk(chunk, len, from);
270 if (err < 0) 274 if (err < 0)
271 goto errout_chunk_free 275 goto errout_chunk_free;
272 276
273 chunk->shkey = shkey; <<
274 <<
275 /* Put the chunk->skb back int 277 /* Put the chunk->skb back into the form expected by send. */
276 __skb_pull(chunk->skb, (__u8 * 278 __skb_pull(chunk->skb, (__u8 *)chunk->chunk_hdr -
277 chunk-> 279 chunk->skb->data);
278 280
279 sctp_datamsg_assign(msg, chunk 281 sctp_datamsg_assign(msg, chunk);
280 list_add_tail(&chunk->frag_lis 282 list_add_tail(&chunk->frag_list, &msg->chunks);
281 } 283 }
282 284
283 return msg; 285 return msg;
284 286
285 errout_chunk_free: 287 errout_chunk_free:
286 sctp_chunk_free(chunk); 288 sctp_chunk_free(chunk);
287 289
288 errout: 290 errout:
289 list_for_each_safe(pos, temp, &msg->ch 291 list_for_each_safe(pos, temp, &msg->chunks) {
290 list_del_init(pos); 292 list_del_init(pos);
291 chunk = list_entry(pos, struct 293 chunk = list_entry(pos, struct sctp_chunk, frag_list);
292 sctp_chunk_free(chunk); 294 sctp_chunk_free(chunk);
293 } 295 }
294 sctp_datamsg_put(msg); 296 sctp_datamsg_put(msg);
295 297
296 return ERR_PTR(err); 298 return ERR_PTR(err);
297 } 299 }
298 300
299 /* Check whether this message has expired. */ 301 /* Check whether this message has expired. */
300 int sctp_chunk_abandoned(struct sctp_chunk *ch 302 int sctp_chunk_abandoned(struct sctp_chunk *chunk)
301 { 303 {
302 if (!chunk->asoc->peer.prsctp_capable) 304 if (!chunk->asoc->peer.prsctp_capable)
303 return 0; 305 return 0;
304 306
305 if (chunk->msg->abandoned) <<
306 return 1; <<
307 <<
308 if (!chunk->has_tsn && <<
309 !(chunk->chunk_hdr->flags & SCTP_D <<
310 return 0; <<
311 <<
312 if (SCTP_PR_TTL_ENABLED(chunk->sinfo.s 307 if (SCTP_PR_TTL_ENABLED(chunk->sinfo.sinfo_flags) &&
313 time_after(jiffies, chunk->msg->ex 308 time_after(jiffies, chunk->msg->expires_at)) {
314 struct sctp_stream_out *stream !! 309 if (chunk->sent_count)
315 SCTP_SO(&chunk->asoc-> <<
316 chunk->sinfo.s <<
317 <<
318 if (chunk->sent_count) { <<
319 chunk->asoc->abandoned 310 chunk->asoc->abandoned_sent[SCTP_PR_INDEX(TTL)]++;
320 streamout->ext->abando !! 311 else
321 } else { <<
322 chunk->asoc->abandoned 312 chunk->asoc->abandoned_unsent[SCTP_PR_INDEX(TTL)]++;
323 streamout->ext->abando <<
324 } <<
325 chunk->msg->abandoned = 1; <<
326 return 1; 313 return 1;
327 } else if (SCTP_PR_RTX_ENABLED(chunk-> 314 } else if (SCTP_PR_RTX_ENABLED(chunk->sinfo.sinfo_flags) &&
328 chunk->sent_count > chunk-> 315 chunk->sent_count > chunk->sinfo.sinfo_timetolive) {
329 struct sctp_stream_out *stream <<
330 SCTP_SO(&chunk->asoc-> <<
331 chunk->sinfo.s <<
332 <<
333 chunk->asoc->abandoned_sent[SC 316 chunk->asoc->abandoned_sent[SCTP_PR_INDEX(RTX)]++;
334 streamout->ext->abandoned_sent <<
335 chunk->msg->abandoned = 1; <<
336 return 1; 317 return 1;
337 } else if (!SCTP_PR_POLICY(chunk->sinf 318 } else if (!SCTP_PR_POLICY(chunk->sinfo.sinfo_flags) &&
338 chunk->msg->expires_at && 319 chunk->msg->expires_at &&
339 time_after(jiffies, chunk-> 320 time_after(jiffies, chunk->msg->expires_at)) {
340 chunk->msg->abandoned = 1; <<
341 return 1; 321 return 1;
342 } 322 }
343 /* PRIO policy is processed by sendmsg 323 /* PRIO policy is processed by sendmsg, not here */
344 324
345 return 0; 325 return 0;
346 } 326 }
347 327
348 /* This chunk (and consequently entire message 328 /* This chunk (and consequently entire message) has failed in its sending. */
349 void sctp_chunk_fail(struct sctp_chunk *chunk, 329 void sctp_chunk_fail(struct sctp_chunk *chunk, int error)
350 { 330 {
351 chunk->msg->send_failed = 1; 331 chunk->msg->send_failed = 1;
352 chunk->msg->send_error = error; 332 chunk->msg->send_error = error;
353 } 333 }
354 334
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.